HAL Id: hal-01940830
https://hal-ifp.archives-ouvertes.fr/hal-01940830
Submitted on 30 Nov 2018
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
O 3 Surfaces
Brice Firmin Ngouana Wakou, Manuel Corral Valero, Pascal Raybaud
To cite this version:
Theoretical insights into the interaction of oxygenated organic molecules
and cobalt(II) precursor with γ-Al
2O
3surfaces.
Brice Firmin Ngouana Wakou, Manuel Corral Valero*, Pascal Raybaud
IFP Energies nouvelles, Direction Catalyse et Séparation, rond-point de l’échangeur de Solaize, BP 3, 69360 Solaize, France
*
Corresponding author: manuel.corral-valero@ifpen.fr
Abstract
We have investigated the interaction of two common oxygenates, ethylene glycol (EG) and acetic acid (AA) with the main exposed (100) and (110) surfaces of γ-Al2O3 surfaces by
quantifying the thermochemistry of the drying process. Using Density Functional Theory (DFT) calculations we have calculated the thermodynamically favorable free energies, and identified the adsorption modes and sites involved for both molecules with the main exposed (100) and (110) surfaces of γ-Al2O3. We show that the affinity of EG for alumina surfaces
sites is stronger than AA after drying. EG preferentially interacts through hydrogen bonds on the hydroxyl nests of the (100) surface, while it may interact through Al-O bonds on the (110) surface. AA is predominantly deprotonated and interacts as acetate anions through Al-O bonds on the (100) surface, and hydrogen bonds on the (110) surface. We then propose a quantitative comparison of the interaction of Co2+ hexaaquo precursor with the same γ-Al2O3
surfaces as found in [Angew. Chem. Int. Ed. 2015, 54, 6824–6827] but also to entropic effects involved in the drying process (vapor release and counter-ion decomposition in the gas phase). We finally discuss how these effects can be circumvented by changing the conditions of the impregnation and drying steps.
1 Introduction
Metal oxides are key materials in heterogeneous catalysis. Amongst them, γ-Al2O3, also
referred to as “alumina”, is one of the most widespread support in industry 1
. Although the main role in a catalytic process is attributed to the so-called active phase (generally a metal), aluminas, acting as “carrier” at the backstage, also have their share of fame in the overall performance of the catalytic material. Once a suitable metallic phase/support couple is identified for a given catalytic reaction, much of the efforts to fine-tune the heterogeneous catalyst focus on the optimization of the concentration, structural (morphology, size) and electronic properties of the metallic active phase, and on the optimization of the textural properties of the support itself. However, the extent to which we may modify any of these characteristics depend on the intrinsic metal support interactions (MSI) between, on the one hand, the forefront active phase and, on the other hand, the alumina carrier at the backstage.
Although many researchers have investigated MSI effect for a long time2–5, the
assessed indirectly from experiment, rationalized with chemical intuition and optimized by trial and error. Even if we focus on the relevant case of alumina supported catalysts, the preparation steps of the catalyst are still scarcely investigated by theoretical approaches 6–8.
During the preparation steps of the catalyst (impregnation and drying), it is well known that the metallic precursors interact with the hydroxyl groups and/or aluminum sites exposed on the aluminas’ surface and this chemical interaction depends on several parameters (pH, counter-ions,…) imposed by the aqueous mother solution4. If we consider relevant
catalytic systems for low temperature Fischer-Tropsch (FT) synthesis 9,10 or for hydrotreating
(HDT) 11, cobalt, nickel and molybdenum are often a key component of the active phases.
Sarrazin et al. have shown by infrared analysis that the interaction of oxomolybdenum anions with alumina surfaces occurs through an exchange of the neutral and basic and hydroxyls
12
. The relationship between the alumina surface structure and the reactivity of surface hydroxyls has been further explored for the adsorption of Mo oxoanions and Ni2+ on both the (0001) and (11̅02) α-alumina monocrystal surfaces13,14
. It was shown that the oxide support is playing different roles depending on the exposed crystal surface. On alumina, DFT calculations have also identified the most favourable adsorption sites of CoII hydrated ions on
the (100) and (110) hydrated surfaces of γ-Al2O3 corresponding to low water pressure after the
drying step and thus in the absence of the liquid solvent 6. In particular, Co anchors on both
γ-Al2O3 (100) and (110) surfaces by substitution of surface OH groups, and by the formation of
additional Co-O bonds with surface oxygen atoms leading to strong epitaxial interactions. An important step further in the preparation methods was made by adding oxygenated organic additives (such as polyols, carboxylic acids, …) to modify and improve the deposition of cobalt metallic precursors of FT and HDT catalysts3. This subject attracts a vivid scientific
understanding of the modes of interaction of such organic additives remains unclear. It is argued that the use of additives might displace the metallic species at the alumina surface and thus reduce the loss of the metal in the support and improve the dispersion or the activation process. These oxygenated organic molecules can be impregnated at different steps of the preparation which impacts the physico-chemical properties of the final FT or HDT catalyst
16,18–21
: on the alumina support before any metallic precursors (pre-impregnation), together with the metallic precursors (co-impregnation), after the metallic precursors and first drying (post-impregnation). In each case, the following pairs of complexation processes are in competition: Co–water, Co–additive, Co–surface, surface–water, and finally surface–additive. Well quantified interactions of all these processes are still missing to tune the choice of additives with the adapted protocol of preparation.
A second important case where the interaction of organic molecules with alumina support plays a key role concerns the preparation of the support itself, and the tuning of its textural properties during hydrothermal treatments 22–24. For instance, alumina supports are often
pre-impregnated (that is, before the metal deposition), with AA in order to increase their porosity. Although this has beneficial effects in some catalytic applications, such as the hydrotreatment of petroleum residues 23, it may be deleterious for the reducibility of the active phase 9,25.
A last important application, where the interactions of metal and organic additives with the support may play a key role, is related to the stability of alumina support during specific reaction conditions, in particular when water is a byproduct of the targeted reaction, such as in FT synthesis or in biomass conversion 26. Several works have shown that adding
metallic elements27 or oxygenated additives 10,28–31 may prevent the transformation of alumina
into boehmite in hydrothermal conditions. It may thus be relevant to better quantify the chemical interaction of such metallic or oxygenated species with the alumina support.
In this theoretical work, we focus our attention on the interactions of acetic acid (AA), ethylene glycol (EG) and compare them with those of cobalt(II) hexaaquo precursors with the adsorption sites of γ-Al2O3. One should bear in mind that the two organic molecules studied in
this work are potential candidates to be used for regulating metal-support interactions during the preparation of alumina supported metal-catalysts.
In the following sections we will first present the theoretical methodology. We then present results on the adsorption of AA and EG for on the two relevant surfaces of alumina, that is, the (110) and (100) surfaces. For that purpose, we will use the relevant alumina surface models from the theoretical works of Digne et al.32,33 that have been already applied to study
MSI 34–39 and alumina-water interface 40 . Then we will compare these results with the case of
cobalt(II) species adsorbed on the same surface to identify how AA and EG molecules can actually impact this cobalt-surface interaction. In particular, we will make use of the results from a recent paper in which the interaction of Co with this support was studied 6. Lastly we
draw the conclusions and present our perspectives for future work.
2 Methodology
The (100) and (110) γ-alumina surface models used in this work are those developed by Digne et al. 32,33. In the conditions of a slight drying between room temperature and 50°C,
both surfaces are considered in their highest hydroxylation state. The structural changes occurring during the hydroxylation of the dry (100) and (110) surfaces of γ-alumina are well explained elsewhere 33,41. There are a lot of hydroxyl groups on these surfaces (Figure 1a) that
offer many different possibilities for the adsorption of the organic molecules such as acetic acid (AA) or ethylene glycol (EG). Note that we will consider not only the adsorption on the hydroxyls but also the exchange of these hydroxyls with the incoming molecules.
2.1 Simulated systems
2.1.1 Adsorption at high surface coverage
Since we have to carry out an extensive exploration of the potential energy surface to find out the most favorable adsorption configurations for these two molecules on both surfaces, it makes sense to start our calculations in the smallest (1x1) simulation cells (Figure 1a) that are obviously less demanding in term of computational resources than the larger cells: (2x3) and (2x2) for (100) and (110) surface respectively. Larger cells are investigated further in our work to explore surface coverage effects (Figures S1.2 and S1.3). Hence we start by studying the adsorption of 1 AA/EG molecule in this smaller cell, which corresponds to surface coverage values θ (100) = 2.15 molecules/nm², and θ (110) = 1.49 molecule/nm²,
Figure 1. (a) Top view of the (100) and (110) hydroxylated slabs of γ-alumina with surface cavities (see Figure S1.2 for the larger systems) indicated for the (100). The aluminum Lewis acid sites (LAS) are numbered by the roman numerals in brackets and their coordination numbers are provided as indices. The surface oxygen and hydrogen atoms of the Brønsted acid sites (BAS) are denoted by letters and arabic numerals respectively. The total number of Al2O3 units is also indicated for each surface, as well as the surface area and the hydroxyl
coverage. Color code: pink (aluminum), red (oxygen), and white (hydrogen). (b) The rugosities of the (100) and (110) surfaces of γ-alumina are compared as a function of the relative positions of the different types of OH surface groups.
hydroxylated (100) and (110) slabs with respective thicknesses of ~20 Å and ~15 Å (referred to as “surface” in Figure 2) and the molecule in solution (referred as “molecule(l)”) were
taken as the reference systems before adsorption. In order to mimic the dry state (gas phase) a vacuum of 25 Å thickness was created on top of the surfaces before and after adsorption (Figures S1.2a and S1.3a). It should be mentioned that the simulation boxes with isolated molecules (in the liquid and gas phases) were created separately for the (100) and (110) slabs in order to keep consistency in the box dimensions of all the sub-systems involved in the adsorption process to avoid errors that may appear due to lateral interactions.
Figure 2. Schematic representation of the impregnation reaction of dried γ-alumina by a solution of acetic acid or ethylene glycol followed by drying at T = 298 K, with x the number of adsorbed AA or EG molecules (1 ≤ x ≤ 6) and y the number of desorbed water molecules during the reaction(0 ≤ y ≤ 6).
Around thousand adsorption configurations were generated for both molecules on the (100) and (110) surfaces to be used as starting points for geometry optimization calculations based on the various schemes provided in Figure 3. In these configurations the molecule can be either dissociated or non-dissociated, and its interaction with the surface can occur through coordinate bonding with Lewis adsorption sites (LAS) indicated by the aluminum atoms in Figure 1a, or through hydrogen bonding with Brønsted adsorption sites (BAS) involving both oxygen and hydrogen atoms of hydroxyls in Figure 1a. One should note that adsorption on LAS AlV or AlVI implies simultaneously the removal of one surface OH/H2O group, whereas
it is not the case for AlIV (to be discussed later). When the organic molecule is linked to the
surface through one of its two oxygen atoms it is referred as to monodentate adsorption. If the two oxygen atoms are implied in molecule-surface connections it is considered as bidentate adsorption. Also if only one surface site (Lewis or Brønsted) interacts with the molecule, then we denominate this configuration either a LAS or a BAS adsorption. Otherwise if there are two or more interacting surface sites, we call them LAS-LAS, BAS-BAS or mixed LAS-BAS adsorption modes. A schematic representation of thecorresponding adsorption configurations is provided in Figures 4a and 4b respectively for AA and EG.
Figure 3. Tree representation of the different adsorption modes and reactions involved during the generation of the starting adsorbed structures (before geometry optimization) of AA and EG on the (100) and (110) of γ-alumina. The adsorption may be dissociative or non-dissociative, and in both situations we have to consider whether a water molecule desorbs from the surface or not, and if the molecule-surface interactions occurs through hydrogen bonds or iono-covalent Al-O bonds.
Figure 4. Schematic representation and denomination of the different initial adsorption modes of AA (a) and EG (b) sorted as a function of the reaction type (dissociative/non dissociative) and adsorption modes (LAS/BAS/LAS-LAS/BAS-BAS/LAS-BAS). Line code: Dashed red represent hydrogen bonds, plain red represent Al-O iono-covalent bonding.
For a given initial adsorbed configuration, several other structures were generated by rotating the molecules around the Al-O bond (LAS) or C-C bond (BAS) as illustrated in Figure S1.1 in the case of AA. When the AA and EG molecules are dissociated upon adsorption, the proton is transferred alternately on each of the surface oxygen atoms and OH groups (Figure 1a) and this creates additional configurations. Note that the proton removed from the organic molecule upon dissociation can also combine with the OH group of the adsorption site to form a H2O molecule that desorbs when the organic molecule binds to
during the generation of our initial adsorbed configurations on the (100) surface, we tried the simultaneous dissociation of the two protons of the OH groups of EG. It appeared that the EG molecule recovers at least one of the dissociated protons after geometry optimization. This observation prevented us from repeating the same procedure on the (110) orientation. Moreover, in all the adsorbed systems generated for acetic acid following the previously mentioned details, two AA molecules were simultaneously adsorbed on the two sides of the slabs on the same adsorption site in order to avoid dipole formation. However, we noticed that the energetic difference between the slabs with induced dipole (that appear due to the adsorption of only one molecule on one side of the slab) and the one with no dipole is negligible (0.002 eV). Hence for EG only one side of the slabs was considered, and the same reasoning was applied for adsorbed systems at lower surface coverages (i.e.: larger supercells) that are discussed hereafter.
2.1.2 Effect of surface coverage
surface coverage values. The objective here is to study how the number of adsorbed molecules affects the adsorption free energies and structures of AA and EG on the two surfaces. For a straightforward comparison of the data obtained at different surface coverages, we keep the same definition of the adsorption energy used before as shown in Figure 2, with N = 1, 2, 3, 4, 6 on the (100) surface and N = 1, 2, 3, 4 for the (110) surface. For all N values our reference state is the hydroxylated γ-alumina surface. Therefore for N > 1 we have a simultaneous adsorption of 2 or more AA/EG molecules on the surface. For simplicity, we consider that for a given adsorption configuration, all the AA/EG molecules are adsorbed at identical adsorption sites on the surface, and through the same adsorption mode and following the same reaction type.
2.1.3 Case of the cobalt system
The calculations carried out in this part were achieved by making use of the large simulations cells as those mentioned in section 2.1.2 for AA and EG. For this purpose, we have reproduced the final cobalt adsorbed configurations reported in Ref.6 on the (100) and
(110) hydroxylated surface models (Figure S1.4) and these structures were further optimized with our set up. In particular, the slabs we used and the geometry optimization procedure differ somehow from Ref. 6 who used a thinner slab. Hence we recalculate the energy values
according to the scheme given in Figure S3.1.
For the free energy calculations, our reference cobalt complex is the charged [Co(H2O)6]2+ in an aqueous solution rather than the neutral gas phase Co(OH)2(H2O)2
complex used in Ref.6, which explains the different energy values The use of this reference
solvated state was necessary to make a coherent comparison with the solvated AA and EG molecules. For that, three separate simulations boxes were created for the simulation of [Co(H2O)6]2+ , H+ and NO3counter-ions. For these systems we used a simulation cubic box
of size 40 Å, as we have noticed that the electronic energies converge (less than 0.04 eV difference) for a box size ≥ 30Å and the implicit solvent model used (see below).
According to thermodynamic data, after drying at 300 K, H2O, N2 and O2 should be
the major products released in the gas phase . However, it is well known that the decomposition of nitrate ions is kinetically hindered. Hence, we decomposed the overall energy change of Co nitrate adsorption into two steps (see Section S3 and Figure 5 below): a) the surface adsorption of Co(II) ions on the surface and b) the decomposition of NO3ions
into N2, O2 and water.
Figure 5: Schematic representation of the impregnation reaction of γ-alumina by a solution of hexaaquacobalt(II) nitrate followed by drying at T = 298 K. The overall reaction is decomposed into: equation (a) the adsorption of the hexaaquacobalt (II) complex and the release of two surface H+ ions (here nitrate ions appears as spectators in both sides of the
equation) and equation (b) the decomposition of NO3- ions to yield water, nitrogen and
oxygen. In (a) “surface” denotes the reference (100) or (110) slabs and “surface-Co(H2O)2” is
the system with the Co(H2O)2 complex placed on the reference surfaces so as to reproduce the
Equation a) allows a comparison for the adsorbate-surface interactions between the organic molecules and the Co precursor, whereas the sum of both equations a and b provides the overall energy change due to the metal impregnation and counter-ion decomposition processes. Moreover, since equation b) is expected to be kinetically hindered, the energy change of equation a and equation a+b could be interpreted, respectively, as the upper and lower limits of the energy change during the impregnation of cobalt nitrate on alumina.
To explore more rapidly the coverage trends, we only performed calculations for the adsorption of a single complex, that corresponds to surface coverage values of θ(100) = 0.36
Co/nm² or θ(110) = 0.38 Co/nm². We determined the difference between our adsorption
energies and the ones reported in Ref. 6 et al. at these θ values, and then we considered the
differential energy to be invariant with respect to θ and based on this assumption, we extrapolated our results to higher θ values. It is important to specify here that for AA/EG molecules, we report the average cumulative adsorption free energy with respect to the number adsorbed AA/EG molecules. Hence to compare cobalt free energies with that of AA and EG we will extract similar data from the incremental free energies of Ref. 6 according the
numerical methods described in supplementary materials S4.
2.2 Energy calculations
The following assumptions were made for the calculation of the adsorption free energies that are discussed in this work:
The vibrational contributions to the adsorption energy compensate between the reactants and the products.
The molecules and the cobalt complex lose 1/3 of their rotational and translational degrees of freedom upon transfer from vacuum to water.
The translational and rotational degrees of freedom of the molecules and cobalt complex in the liquid phase are converted into bound motions after adsorption.
The adsorption free energies of the organic molecules and the cobalt complex were calculated as described in S3 section of the supplementary electronic materials. The electronic energies (Eel terms in S3) were calculated from geometry optimizations using periodic density functional calculations with the VASP code 42–45 and the Generalized Gradient Approximation
PBE functional (GGA-PBE) 46,47 within the projected augmented-wave (PAW) method
48,49
.The solvation effects for our systems in the liquid phase (Figures 2 and 6) were treated with an implicit solvation model (we used the dielectric constant of water) that is implemented in the VASPsol software 50. Dispersion corrections were included in our
calculations through the Grimme-D2 method 51. For all our simulated systems, the geometry
were considered to be optimized when the ionic forces were smaller than 0.02 eV/Å and spin polarized calculations were performed for the systems with cobalt. For a rapid screening of the 1000 initial configurations in section 2.1.1 we used an energy cutoff of 300 eV for the plane-wave basis set and a normal precision. Later on a high precision and a cutoff of 400 eV were used for the calculations in sections 2.1.2 and 2.1.3. Integration over the Brillouin zone was done in a set of k-points provided by the Monkhorst-Pack algorithm 52 within a grid of
0.02 Å-1 fineness.
3 Results
3.1 Adsorption at high surface coverage
3.1.1 Energetics
(PES)- with those obtained with larger slabs –larger than the unit cell- at the same surface adsorbate coverage. Adsorption energy values of AA and EG molecules are reported in Figure 6. From these data the most favorable adsorption configurations (lowest G) were refined using larger simulations cells as explained before, and the refined structures and energies are reported in Figures 7 and 8. The free energy data indicate a non-systematic trend regarding the surface size: the energy values are slightly smaller for AA in the large systems for both surfaces, and for EG on the larger (110) surface. However EG free energy values are slightly higher on the larger (100) surface, compared to the smaller systems. These fluctuations do not exceed 10 kJ/mol, and can be attributed to the H-bonds stretching or rather, to a small reorganization of H-bonds on the γ-alumina surfaces. In the following sections of this manuscript, the free energy values that are presented and discussed, as well as the adsorption configurations, are those obtained in the large cells.
(b), EG on the (100) surface (c) and EG on the (110) surface (d). The filled symbols represent the free energy values of the most favorable configurations refined in the large cells.
When adsorbed, AA binds more strongly on the (100) surface than on the (110) surface as shown by the most exergonic ΔG values (for each plot in Figures 6a and 6b). The most favorable adsorption configuration on the (100) surface has a free energy value that is more than twice its equivalent on the (110) surface (ΔG(100) ≈ -88 kJ/mol while ΔG(110) ≈ -37
kJ/mol). Moreover, the two other favorable adsorption configurations on the (100) surface have free energy values that are also more exergonic than those for any of the most favorable adsorption modes on the (110) surface (Figures 6a and 6b). As for EG, similarly to AA, the three most favorable adsorption configurations of this molecule on the (100) surface exhibit more negative ΔG values than those on the (110) surface (Figures 6c and 6d). Moreover the most favorable adsorption configuration of EG on the (100) surface has a free energy value almost three times higher than his counterpart on the (110) surface (ΔG(100) ≈ -132 kJ/mol
while ΔG(110) ≈ -47 kJ/mol). It is clear from these observations that AA and EG molecules
have a significantly higher affinity towards the (100) surface in comparison with the (110) surface. Moreover the free energy differences between the two surfaces are greater for EG than for AA.
3.1.2 Structural features of the adsorption configurations
a) Adsorption modes
For both surfaces, the most favorable configurations of AA belong to the bidentate BAS-BAS adsorption mode, here denoted asas AA1(100) and AA
(110)
1 (Figures 7b and 8b). The
here denoted as AA2(100) and AA2(110) (Figures 7b and 8b). This requires not only the dissociation of the AA molecule, but also the removal of one H2O molecule from the (100)
surface to make available one AlV site for the anchoring of AA. On the (110) surface, AA can
directly bind the AlIV site without water removal (Figure 8b). AA2(100) and AA2(110) are clearly
less favorable than AA1(100) and AA1(110), respectively (Figures 7b and 8b). On the (100) surface, a third favorable adsorption configuration, called AA3(100), belongs to the monodentate BAS-BAS adsorption mode (Figure 7b) and is energetically very close to AA2(100) (-47 kJ/mol vs -54 kJ/mol). In this case the AA molecule does not dissociate upon
Figure 7. Molecular views of the most favorable adsorption configurations of AA and EG on the (100) surface in the large cells at θ(100) = 2.15 molecules/nm². Subfigures (a) and (c) are
Figure 8. Molecular views of the most favorable adsorption configurations of AA and EG on the (110) surface in the large cells at θ(110) = 1.49 molecules/nm². Subfigures (a) and (c) are
Similarly to AA, EG binds more strongly to the (100) surface compared to the (110) one, as revealed by the most negative ΔG values in Figures 6c and 6d. It also appears from these figures that the EG molecule prefers in most cases to form hydrogen bonds with both surfaces as its most stable configurations (EG(100)1 and EG(110)1 ) belong to the bidentate BAS-BAS mode (see Figures 7d and 8d). Unlike AA, no proton dissociation is involved in the adsorption process leading to EG(100) 1 and EG(110) 1 configurations. Furthermore, two other favorable adsorption configurations are identified for each surface, all belonging to the bidentate LAS-BAS mode (Figures 7d and 8d). In both cases EG forms a covalent bond with the surface and a H2O molecule is released. It is interesting to see that EG can form such a
iono-covalent Al-O bond with the (100) and (110) surfaces of γ-alumina following either a dissociative mode (see EG(100) 2 in Figure 8d and EG(110) 2 in Figure 8d) or a non-dissociative one (see EG(100) 3 in Figure 7d and EG(110) 3 in Figure 8d). Our energetic analysis shows that these two modes, where EG forms iono-covalent Al-O bond, are twice as less favored than the BAS-BAS EG1 mode on the (100) surface, whereas on the (110) the three modes are closer in
energy. This also means that the Al site on the (100) prefers to remain coordinated to one water molecule rather than to EG. This result is consistent with FT-IR28 and Sum Frequency
Generation (SFG)53 showing that EG interact predominantly through hydrogen bonds on
alumina surfaces. in presence of water at RT. In the present study, the non-negligible hydroxyl groups present on the alumina surfaces at T=298 K explains also this trend. Moreover, we show that this trend may be more pronounced on the (100) surface than on the (110).
When analyzing the nature and the number of the interacting hydroxyl groups for the AA1(100) , AA1(110) , EG(100) 1 and EG(110) 1 configurations, we count 6 and 4 interacting hydroxyls for each molecule respectively on the (100) surface and the (110) surface. For AA, these interacting hydroxyls are 4 H2O-μ1, 1 HO-μ2, and 1 HO-μ3 surface groups on the (100)
surface (Figure 7b) and 1 H2O-μ1, 2 HO-μ1, and 1 HO-μ2 surface groups on the (110) surface
(Figure 8b). Considering this analysis, the (100) surface involves a higher number of H-bonds (hydrogen bonds are defined according to IUPAC recommendations of 2011 54; we do not
consider any hydrogen bond longer than 3.0Å) and also surface hydroxyls such as 1 HO-μ3
and H2O-μ1 which are stronger H-bond donors (ie more acidic) than on the (110) surface (with
2 HO-μ1), this may explain the more exergonic G values found on the (100) surface. For
EG, we identify 1 H2O-μ1, 4 HO-μ1, and 1 HO-μ2 surface groups on the (100) surface (Figure
8d) and 2 H2O-μ1and 2 HO-μ1 surface groups on the (110) surface (Figure 8d).
For the AA3(100) configuration (Figure 7b), for which no proton dissociation is involved in the adsorption process and where AA is also interacting with the surface only through hydrogen bonds, there are 3 interacting surface groups: H2O-μ1, HO-μ1, and HO-μ3. As
previously described, the AA2(100) and AA2(110) configurations reveal that the molecules interact through iono-covalent O-Al bond formed between AA and the AlV site of the (100)
surface (dO-Al = 1.99 Å, Figure 7b) and the AlIV site of the (110) surface (dO-Al = 2.03 Å,
Figure 8b). Moreover, both AA2 structures interact with 5 sites on the (100) surface (2 H2
O-μ1, 1 HO-μ1, 1 HO-μ2, and 1 HO-μ3), and only 3 sites on the (110) surface (1 H2O-μ1, 1
HO-μ1, and 1 HO-μ2). As for AA1, the nature and number of H-bond donors involved explain
easily the energetic trends. Moreover, the Al-O bond length is slightly shorter on the (100) which may reinforce the interaction on this surface. In the EG(100) 2 configuration, the molecule interacts with 2 H2O-Al, 3 HO-μ3 and 1 O-μ3 groups on the (100) surface with a covalent
Al bond length of dO-Al = 1.95 Å (Figure 7d). For the analogous EG(110) 2 configuration on the
(110) surface, only one H2O-μ1, one HO-μ1, one HO-μ3 H-donor and one O-μ3 bond acceptor
are the surface interacting sites and dO-Al = 1.84 Å (Figure 7d). Unlike AA, one possible mode
(although less energetically favorable) of EG is the non-dissociative bidentate LAS-BAS adsorption found on the two studied surfaces (EG(100) 3 and EG(110) 3 ). The EG(100) 3 and EG(110) 3 configurations involve 4 interacting surface groups on the (100) surface (H2O-Al, HO-μ1, HO-μ3 and O-μ3, see Figure 7d), and on the (110) surface (H2O-Al, HO-μ2,
O-μ2 and O-μ3, Figure 8d). The small energetic differences appearing between
EG(100) 2 and EG (100) 3 (ΔΔG
(100) = 3 kJ/mol, see Figure 6d) and between EG(110) 2 and EG(110) 3
(ΔΔG(110) = 7 kJ/mol, see Figure 7d), can be attributed to the O-Al covalent bond interactions
of EG(100) 3 and EG(110) 3 that are weaker than those for the EG(100) 2 and EG(110) 2 configurations. Moreover these tiny energy differences may also indicate that the protons marked by an arrow in Figures 6d, 7d, S2.2d and S2.3d are prone to dissociation, and this proton dissociation may strengthen the covalent O-Al bond interaction and yield EG(100) 3 and EG(110) 3 configurations as stable as those in EG(100) 2 and EG(110) 2 .
3.1.3 Effect of the alumina facet on the adsorption structures and energies
We now focus on the interaction of AA and EG with different surface facets. Hereafter, surface H2O-Al and HO-μ1 groups above the average surface plane (see figure 1b)
we see that the external H2O-Al and HO-μ1 groups interact with AA and EG molecules on the
(100) and (110) surfaces while the internal HO-μ2, and HO-μ3 groups are found in the case of
the (100) surface only. Note that the HO-μ2 group interacting with AA on the (110) surface is
an external surface group, according to our definition in Figure 1b). It is clear from these observations that the strength of molecule-surface interactions and the corresponding free energy differences that we found for both molecules on each of the two surfaces can be attributed not only to the nature and number of interacting sites, but also to the surface topology or roughness. On the (100) surface AA and EG molecules are able to insert in the surface cavities (Figures 1a and 7a and 7c) and to bind to the hydroxyl nests surrounding these cavities. If we evaluate the position of EG with respect to the average surface plane: it is clearly shifted downwards on the (100) surface (Figure S2.1b) with respect to the (110) surface (Figure S2.1d), and this also enhances significantly the molecule-surface interaction.
On the (110) orientation, surface roughness (Figure 1a) prevents the AA and EG molecules to closely contact the surface such that there is a smaller number of interacting hydroxyls accessible to the molecules (Figure 8 and Figure S2.1c). A similar effect was invoked by the study of xylitol adsorption on boehmite surfaces 55, where the presence of such
hydroxyl nests play a key role in stabilizing the xylitol on the surfaces. Therefore, the stabilization may depend both on the type of the facets and on the corresponding topology of the hydroxyl cavities.
The AA2(100) , AA2(110) , EG(100) 2 , EG(110) 2 , EG(100) 3 , and EG(110)3 configurations are characterized by the O-Al covalent bond between the molecule and the surface. We recall that in the AA2(110) configuration there is no H2O desorption involved, as it is the case for all other
moderate. This can be ensured by desorbing a H2O molecule from the surface. Besides, when
this H2O is taken out from the (100) surface, AA and EG in the AA2(100) , EG(100) 2 ,
and EG(100) 3 configurations have more room to interact with internal HO-μ3 and even O-μ3
surface groups (Figures 7b and 7d). Interestingly on the (110) surface, the water molecule is so strongly held on the AlIV site that the energy cost for breaking a H2O-Al bond (and the
H-bonds connections with it) on the surface is not compensated by the formation of the new O-Al bond between AA and the (110) surface. Hence the only way to become exergonic (-9 kJ/mol) upon adsorption of AA is to keep a H2O molecule on the AlIV site of the (110)
surface. In fact, when the H2O molecule is removed from the AlIV before the adsorption of
AA, the adsorption free energy has a maximum value of +4 kJ/mol. By contrast, EG bonds with the (110) orientation are strong enough to fully compensate the energy cost needed to take out 1 H2O from the surface, so that more energy is gained in the process: -9 kJ/mol for
AA2 (Figure 8b) compared to -28 for EG2 or -35 kJ/mol for EG3 (see Figure 8d). One should
note that in the EG2 and EG3 configurations the EG molecule is covalently bonded to the
AlVI(II) and AlVI(III) sites on the (110) surface (Figure 1a), respectively. This situation can
only happen if a water molecule is removed from these sites. This observation clearly indicates that for a given surface, the nature of the adsorbed molecule significantly affects molecule-surface interactions.
3.1.4 Influence of the molecule on the adsorption energies and structures
Figures 7-8 show that both surfaces have a higher affinity for EG than for AA, and the energetic trends between AA and EG also vary according to the adsorption mode.
According to Figures 7b and 7d, the free energy change for the EG1(100) structure is
more exothermic than the one for the AA1(100) structure by 44 kJ/mol, while the difference
between EG(110)1 and AA (110)
1 structures is 10 kJ/mol (Figures 9b and 9d). The calculated free
energy values indicate that, when only hydrogen bonds connect the molecule to the surface, AA and EG interacts very similarly with the (110) surface. This assertion is reinforced by the same number of hydrogen bonds (4 in each case) that relates both molecules to the (110) surface (Figures 8b and 8d). Oppositely, on the (100) surface the bidentate BAS-BAS configurations of AA and EG have respectively 6 (Figure 7b) and 7 (Figure 7d) hydrogen bonds with the surface. Their corresponding free energy values indicate that EG binds more strongly than AA on the (100) surface. First, the slightly longer and flexible O-C-C-O chain of EG is more suitable, compared to AA, for occupying two (100) surface cavities as discussed above so that it provides hydrogen bonds with almost all the surface hydroxyl groups (external and internal) found around those cavities. EG is thus more prone to be oriented in a parallel position with a rather symmetric distribution of H-bonds involving the two OH groups of EG, whereas AA is tilted with a non-symmetric distribution of H-bonds on the two O atoms (Figure S2.1). Moreover, according to the distributions and orientations of the two molecules on the (100) surface (Figures 7a, and 7c) and the (110) surface (Figures 8a, and 8c), molecule-molecule interactions are suspected to be stronger on the former surface than on the later. The molecule-molecule and molecule-surface stabilizing dispersion forces contribution will favor EG adsorption compared to AA, and this effect seems to be more significant on the (100) surface than on the (110) surface. All these parameters explain the 44 kJ/mol energy difference between AA and EG on the (100) surface.
On the (110) surface, due to the smaller energy difference between AA and EG, it is far more difficult to extract one unambiguous parameter differentiating the behavior of the two molecules.
When covalent bonds are formed between the molecules and the surfaces, adsorption free energies of both molecules differ by ΔG(AA2(100)) - ΔG(EG
(100)
2 ) = 13 kJ/mol on the (100)
1.95 Å (Figure 7d) while AA has 5 hydrogen bonds and one slightly longer O-Al bond distance ( 1.99 Å) (Figure 7b). On the (110) surface, ΔG(AA2(110)) - ΔG(EG(110)2 ) is about + 26 kJ/mol. However, as mentioned above, the adsorption process of AA2 does not involve water
removal, so there is no entropic compensation as it is the case for EG. Thus, although ΔU(AA2(110)) is more exothermic than ΔU(EG(110)2 , it becomes the reverse for the ΔG values due to entropic effect. Hence, when a covalent bond is formed with the surface, EG adsorption free energies on both surfaces are stronger compared to AA, and this effect is more pronounced on the (110) surface compared to the (100) surface due to entropic effect of water. Moreover EG being more flexible than AA, the molecule is able to deform its O-C-C-O backbone easily in order to maximize the interactions with the surface. EG exhibits many flat lying configurations where its C-C bond is oriented parallel to the surface, whereas AA is often oriented perpendicularly or tilted with respect to the surface.
3.2 Effect of surface coverage
To better control the optimum surface loadings to be used during the impregnation step of the γ-alumina surfaces by organic additives, it is important to know how the number of molecules adsorbed on the surface affects the structure and energy of adsorbed configurations. The results presented and discussed above were obtained at surface coverage values of θ (100)
= 2.15 molecules/nm² and θ (110) = 1.49 molecules/nm² ( the highest values considered for both
Figure 9. Adsorption free energy as a function of surface coverage on the (100) surface for AA (a) and for EG (c), and on the (110) surface for AA (b) and EG (d). In each case the cobalt adsorption free energy values were calculated in this work at the lowest coverage and the other values were exptrapolated based on the results of Ref. 6. For each plot, the triangles, circles and diamonds are the respective free energy values of each configuration calculated in the small cells at θ(110) = 1.49 molecules/nm² and θ(110) = 1.49 molecules/nm².
The trends in surface coverage effects are the outcome of a trade-off between, on the one hand, surface-molecule and molecule-molecule interactions of adsorbates and, on the other hand, disrupted interactions between surface OH groups upon adsorption. The analysis of adsorption sites involved at θ (100) = 0.36 AA/nm² (Figure S2.2b) and at θ(100) = 2.15 AA/nm²
(Figure 7b) shows that they are very similar in both cases, except for the AA2(100) configuration for which the molecule is in interaction through a hydrogen bond with an additional μ1-OH group at θ(100) = 0.36 AA/nm². This μ1-OH (see the arrow on Figure S2.2a,
and S2.2b) is actually the one that is attached to the AlV sites, and which is desorbed (in the
form of a H2O molecule after combining with the AA dissociated proton) prior to AA
adsorption on that Alv site. Hence as θ decreases more of these OH groups become available
to interact for a given AA molecule present on the surface. It is also clear from Figures 7b and S2.1b that the hydrogen bonds and covalent bonds distances do not change dramatically for different θ values. It follows from these observations that the slight free energy fluctuations that occur upon changing the θ value, do not appear to be directly influenced by the local structural changes around the adsorbed molecules, but are rather due to lateral interactions: surface molecule-molecule interaction and hydrogen bond networks disruption. In particular, in the AA3(100) configuration, there are only 3 interaction sites attaching AA to the surface (Figures 7b and S2.1b). In this case the molecule-surface interactions created on the surface by the incoming molecule do not compensate the hydroxyl-hydroxyl interactions loss which bring a more important stabilizing contribution at θ(100) = 0.36 AA/nm² than at θ(100) = 2.15
AA/nm². By contrast, AA1(100) exhibits twice as many interacting sites (6) on the surface (Figures 7b and S2.2b). This significantly increases the contribution from molecule-surface interactions at θ(100) = 2.15 AA/nm² and reduces the energy difference with that at θ (100) =
0.36 AA/nm² (Figures 7b and S2.2b). For the AA2(100) configuration, there are also many interaction sites on the surface to enhance molecule-surface interactions contributions at θ (100)
= 2.15 AA/nm². However, water molecules are desorbed from the surface during the adsorption process, and the energy cost associated with the multiple desorption reactions is higher than the stabilizing contribution arising from the numerous molecule-surface and molecule-molecule interactions at θ (100) = 2.15 AA/nm².
Ethylene glycol
The adsorption reaction of EG molecules on the (100) surface is exergonic at all θ values, except for the EG(100)3 configuration at θ (100) = 0.72 EG/nm², and for the EG(100)2
configuration at θ (100) = 1.08 EG/nm² (Figure 9c). Two different trends are observed for EG:
when θ increases, the adsorption free energy decreases for the EG(100)1 configuration (hydrogen bonded) whereas it fluctuates for the EG(100)2 , and EG(100)3 configurations (covalently bonded)). This trend thus differs from the AA molecule. However, the EG(100)1 configuration remains the most stable at all θ values studied in this work.
According to our structural analysis, unlike AA, the nature of surface interacting hydroxyls with EG are notably changed between the lowest and highest θ values, which explains also why the energy values depend more strongly on the coverage than for AA (fluctuations up to 70 kJ/mol for EG vs 40 kJ/mol for AA). As previously mentioned and as illustrated in Figure 7a and 7c, the C-C backbone of EG lies parallel to the surface whereas the one of AA is perpendicular or tilted. It appears from Figure 7c and S2.2c that for the EG1(100) configuration, the contribution coming from the growing number of molecule-surface
interactions at θ (100) = 2.15 EG/nm² is weaker than the contribution from surface-surface
interactions disrupted from the surface coverage at θ(100) = 0.36 EG/nm². Moreover, as we can
see in Figure 6c, at θ (100) = 2.15 EG/nm² the EG molecules are not in close contact with each
other and this reduces the contribution from molecule-molecule interactions to the free energy. As a consequence, the EG adsorption on the (100) surface in the EG(100)1 configuration is more favorable at low θ values. For the EG(100)2 , and EG(100)3 configurations, we can see in Figure 7c that the orientation of the adsorbed EG molecules on the surface encourages them to interact with each other at θ (100) = 2.15 EG/nm². As a result the
that, it not only compensates for the energy cost of disrupting the multiple H2O desorption
reactions at θ (100) = 2.15 EG/nm², but also stabilizes the adsorbed configurations. In addition,
for the EG(100)3 configuration, there are hydrogens bonds between the EG molecules at θ (100) =
2.15 EG/nm² (Figure 6c). This reinforces the stability of that configuration as the energy gain we observe between θ (100) = 0.36 EG/nm² and θ (100) = 2.15 EG/nm² is more important in
comparison to that of the EG(100)2 configuration (Figure 7d and S2.2d). 3.2.2 (110) surface
Acetic acid
Contrary to the (100) surface, the adsorption free energy of AA on the (110) surface becomes more exergonic with increasing surface coverage for both AA1(110), and AA2(110) configurations (Figure 9b). The AA1(110) configuration is always more stable than the AA2(110) configuration. However, the energies remain significantly lower than on the (100) in both cases. Moreover, the impact of the coverage is weak and it is coherent with the fact that no significant local structural changes occur around the AA molecule (Figures 8b and S2.3b). As for the (100) surface, since the C-C backbone of the AA molecule is nearly perpendicular to surface, the lateral interactions are expected to be small as illustrated in Figure 8a. A weak molecule-molecule interaction may contribute to the stabilization of the system with increasing coverage.
Ethylene glycol
for the EG(110)1 than for the EG(110)2 , and EG(110)3 configurations. Moreover, EG1(110) becomes the most stable configuration for θ (110) ≥ 0.74 EG/nm² whereas it is the less stable
configuration when θ (110) = 0.37 EG/nm².
3.3 Comparison between oxygenates and cobalt(II)
3.3.1 Energetic analysis
Knowing the structural features and adsorption free energy values of the most favorable adsorption configurations of AA and EG on the (100) and (110) surfaces of γ-alumina, we now compare these data with those deduced from the adsorbed cobalt complex Co(H2O)2 previously studied in reference 6 to identify the possible competition between
organic-surface interactions and metal-surface interactions.
First of all, we point out that the formation of Co(H2O)2 complexes on the (100) surface is
more favorable than on the (110) surface, whatever the coverages (Figure 9). This trend is thus similar as for the adsorption of AA and EG molecules. .
The analysis of the adsorption of the Cobalt species on both alumina surfaces (after drying) reported in Figure 9 indicates that the overall process (the sum of equations a and b in Figure 5) is an exergonic reaction, and the corresponding free energy values are always far more exergonic than those of AA and EG. On the (100) surface (Figures 9a and 9b) the adsorption of a 2nd cobalt complex is slightly more favorable than that of the first 1st one,
unlike what we observe for almost all AA and EG configurations. Regarding the (110) surface, the Co(H2O)2 adsorption free energy becomes slightly less exergonic with increasing
θ value, whereas that of AA and EG increase. Nevertheless, at θ(110) = 1.49 Co(H2O)2/ nm² the
We must point out here that the ΔG values reported in Figure 9 corresponds to the impregnation of the metal salt followed by the thermal decomposition of the nitrate counter-ion into (N2 + 5/2O2) after drying. The reaction free energyof nitrate decomposition (equation
b in Figure 5)is -206 kJ/mol at the considered drying conditions (T=298 K). As a consequence the free energy for the adsorption of Co(H2O)2 complexes on the (100) and (110) surfaces are
thus far less exergonic : -79 and 0 kJ/mol, respectively (equation a in Figure 5). The latter ΔG
values do not depend on the choice of the counter-ions, and stand for the intrinsic process of the formation of the dried surface Co(H2O)2 complexes(without involving the counter-ions
decomposition) leading to 4 water molecules released in gas phase and the two solvated protons. The corresponding internal energies of the adsorption of Co(H2O)2 complexes (after
drying) are ∆U(100)Cobalt = +62 and ∆U(110)Cobalt = +141 kJ/mol, which highlights the entropic effect induced by the release of the 4 H2O moleculesin gas phase (see also section S3 of
supplementary materials) at 298K. This implies thus a much larger entropic contribution in favor of Co(H2O)2 compared to AA and EG for which zero or one H2O molecule is released
in the gas phase upon adsorption and after drying. This entropic effect is amplified by the simultaneous decomposition of the nitrate counter-ions. Hence, for the drying process explored so far, the free energy balance is clearly in favor of the cobalt precursor.
become less favorable to cobalt complexes formation. Lastly, if we qualitatively consider the conditions reached at the impregnation step of (before the drying step), all entropic effects due to the final state are significantly reduced. The G values of adsorbed Co(H2O)2 complexes
would become closer to the previous estimates (-79 kJ/mol and 0 kJ/mol) following equation (a) in Figure 15. These G values become less favorable with respect to those calculated for AA and EG molecules. Although this quantitative analysis requires dedicated calculations dealing with the solvated adsorbed species (beyond the scope of the present work), we suspect from this qualitative analysis that the organic molecule may compete more easily with Co(H2O)2 complexes for the surface during impregnation steps (in the solvated state). If so,
the displacement of the strongly interacting Co(H2O)2 complexes by the molecules will be
enhanced if kinetic limitations are involved during the drying process.
3.3.2 Structural analysis of adsorption site competition
Figures 10 and 11 summarize the locations of Co(H2O)2, AA and EG according to
their most favorable adsorption configurations at the lowest coverages studied here on the (100) and (110) surfaces, respectively. Regarding the cobalt species, in order to better depict the effect of surface coverage, we also represent schematically the various cobalt complexes in epitaxial interactions with both surfaces at high cobalt coverages according to the previous proposals 6. On the (100) surface, the first adsorbed cobalt complex sits at the edge of a
(100) surface, the Co(H2O)2 complexes preferentially sit in an epitaxial configuration, and the
subsequent Co(H2O)2 complexes reside also at the edge of a (100) surface cavity. This leads
to two possible epitaxial relationships either along the x-direction or the y-direction of the cell. Hence, we can suggest that if AA and EG molecules are pre-impregnated on the surface before cobalt adsorption, they might break the epitaxial relationship involving some Co(H2O)2
complexes particularly those located along the x-direction (light blue square). However, the complete hindering of Cobalt complexes could be possible only if the surface coverages of EG and AA must reach 2.15 molecule per nm2 at least so that all cavities are occupied and
cannot interact anymore with Co(H2O)2 complexes. . Moreover, considering their main
Figure 10. Top views of the (100) surface showing the location of the adsorbed species: Cobalt, acetic acid, and ethylene glycol at the lowest surface coverage (0.36 species per nm2) The blue squares represent schematically the two directions for the epitaxial relationships of the Co complexes at higher coverages (light blue: 6.70 Co atoms per nm2, dark blue: 4.47 Co atoms per nm2) according to the schemes proposed in Ref. 6. The adsorbed species and interacting OH groups are colored as follows: grey (carbon), red (oxygen), white (hydrogen), and blue (cobalt). The dark green color is used for all other surface atoms.
Figure 11 reports the scheme proposed by these authors who found an epitaxial relationship along the y-direction. According to this representation, EG(110)2 is a configuration prone to hinder the epitaxy because it is located on the same row where the cobalt chains may grow. So from a structural point of view, the molecules are able to compete with cobalt for the anchoring sites and we cannot exclude that the EG and AA may impact the interaction of cobalt species with this alumina (110) surface particularly in the AA1(110) and EG(110)2 configurations.
As previously discussed, from the energetic point of view, the competition is thermodynamically strongly in favor of cobalt-surface interaction after drying at RT. So, one should attempt to compensate it by using a sufficient concentration of organic compounds during a pre-impregnation step of the surface to pre-cover the surface the anchoring sites of cobalt. As for the (100) surface, one may assume that the epitaxial growth can only be prevented by AA and EG molecules at the early stage of the cobalt impregnation step once the alumina surface has been impregnated by the organic molecules and when entropic effects are minimized by the solvent. After drying, cobalt should displace both EG/AA molecules according to the thermodynamic driving forces. If one assume that kinetic limitations will prevent the displacement of the organic molecules by cobalt complexes during drying, the effect of EG/AA can be also effective after drying. This aspect remains to be investigated by further theoretical calculations.
Figure 11. Top views of the (110) surface showing the location of the adsorbed species (Cobalt, acetic acid, and ethylene glycol) at the lowest surface coverage (0.37 species per nm2) The blue losanges represent schematically the epitaxial interaction of the Co complexes at higher coverages (4.44 Co atoms/Å2), according to the schemes proposed in Ref. 6. The interacting surface sites and the species are highlighted with the following colors: grey (carbon), red (oxygen), white (hydrogen), and blue (cobalt). The dark green coloris used for all other surface atoms.
4. Conclusions
Using periodic DFT calculations, the potential energy surface of two oxygenated organic additives, namely acetic acid (AA) and ethylene glycol (EG) adsorbed on the (100) and (110) surfaces of γ-alumina, have been explored. These molecules are potential candidates for tuning cobalt-surface interactions during catalysts preparation. The interactions of these molecules with the above mentioned surfaces were studied in order to find their preferred adsorption modes and adsorption sites, as well as their adsorption free energies. Several adsorption modes were then identified for both molecules on the two surfaces and, for both AA and EG, the most favorable configurations on a given surface are Brønsted adsorption modes involving only hydrogen bonds connections between the molecule and the surface. In particular, a strong interaction of EG with the hydroxyls nests (cavities) of the (100) surface has been found. The molecules can also bind to the surfaces through a covalent bond in a Lewis adsorption mode, but the obtained structures are less energetically favorable than the previous ones. In fact for a given molecule, the comparison of the adsorption free energies of its most probable adsorption configurations indicate that AA is likely to form a covalent bond with the two γ-alumina surfaces, whereas EG would most probably do so only on the (110) surface. In general both molecules have a much higher affinity for the (100) surface than to the (110) surface, and both surfaces have a higher affinity for EG than for AA.
surface and with the nature of the adsorption configuration. On the (100) surface, the adsorption free energies of the three AA configurations decrease with increasing θ value, while on the (110) surface the adsorption free energies increase as surface coverage increases. For EG on the (100) surface, if covalent bonds are involved in EG-surface interactions (LAS), high θ values are more favorable. However if only hydrogen bonds link the molecule to the surface (BAS), low θ values are then preferred. On the (110) surface, the adsorption free energies of the EG configurations with a covalent bond are almost invariant, while the free energy of structures with only hydrogen bonds with the surface lowers with decreasing θ value. In all cases our data show that the structures of adsorption configurations are hardly affected by the changes in surface coverages, so that the observed energy changes are only related to, on the one hand, the number and the type of surface-molecule and molecule-molecule interactions and, on the other hand, the extent to which surface OH groups are disrupted by the incoming molecule.
A comparison with the adsorption of the Co(H2O)2 complex has been undertaken to
explore if the surface-molecule interactions can help us to predict the influence of these molecules on surface-Co(H2O)2 interactions. As for AA and EG, the Co(H2O)2 complex have
a higher affinity for the (100) surface relatively to the (110) surface. However, after drying, both surfaces showed a significantly higher affinity for the Co(H2O)2 complex than for any of
the two molecules and for both surfaces the affinity order is: Co(H2O)2 >> EG > AA. From
the analyses of the surface locations of the organic species, and by comparing them with cobalt epitaxial structures proposed in a previous work 6, we have identified some relevant
configurations where two molecules may compete for the same adsorption sites as the ones involved in Co(H2O)2 adsorption. However, since the thermodynamic analysis (after drying)
impregnation step when the entropic contribution of the gas phase products formed upon drying is minimized. More particularly, these molecules can eventually be efficient if two consecutive impregnation steps of alumina are used: the first one involving the organic molecules only and the second one the cobalt precursors assuming that kinetic limitations may also take place during drying. In order to validate this concept we may propose as a challenging perspective of the present work to investigate the kinetic aspects of the adsorption and desorption processes in presence of an implicit representation of the aqueous solvent. Hopefully, the present study may already offer relevant insights for a more rational understanding of the preparation steps of alumina supported catalysts and may open the route for further investigations on different organic molecules and metallic precursors.
ASSOCIATED CONTENT
Supporting Information.
S1. Details of our simulated systems
S2. Snapshots of acetic acid and ethylene glycol configurations at the lowest surface coverage
S3. Calculation of the adsorption free energies
S4: Extrapolation of adsorption data from Larmier et al. 6
This material is available free of charge via the Internet at http://pubs.acs.org.
Manuel Corral Valero, manuel.corral-valero@ifpen.fr
Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.
Acknowledgements
This work was supported by the French National Research Agency within the framework of the ANR-14-CE08-0019 SLIMCAT project. This work was performed using HPC resources from GENCI-CINES (Grant A0020807386 ).
References
(1) Euzen, P.; Raybaud, P.; Krokidis, X.; Toulhoat, H.; Le Loarer, J. L.; Jolivet, J. P.; Froidefond, C. Alumina. Handbook of Porous Solids; Wiley-VCH Verlag GmbH, 2008; pp 1591–1677.
(2) Mitchell, S.; Michels, N. L.; Pérez-Ramirez, J. From powder to technical body: the undervalued science of catalyst scale up. Chem. Soc. Rev. 2013, 42, 6094–6112.
(3) Munnik, P.; de Jongh, Petra E; de Jong, Krijn P. Recent Developments in the Synthesis of Supported Catalysts. Chem. Rev. 2015, 115, 6687–6718.
(4) Che, M. Interfacial Coordination Chemistry - Concepts and Relevance to Catalysis Phenomena. Stud. Surf. Sci. Catal. 1993, 75, 31–68.
(5) Bourikas, K.; Kordulis, C.; Lycourghiotis, A. The Role of the Liquid‐Solid Interface in the Preparation of Supported Catalysts. Catal. Rev. 2006, 48, 363–444.
(6) Larmier, K.; Chizallet, C.; Raybaud, P. Tuning the Metal–Support Interaction by Structural Recognition of Cobalt-Based Catalyst Precursors. Angew. Chem. Int. Ed. 2015, 54, 6824– 6827.
(7) Digne, M.; Raybaud, P.; Sautet, P.; Guillaume, D.; Toulhoat, H. Quantum chemical and vibrational investigation of sodium exchanged [gamma]-alumina surfaces. Phys. Chem.
(8) Louwerse, M. J.; Rothenberg, G. Modeling Catalyst Preparation: The Structure of
Impregnated–Dried Copper Chloride on γ-Alumina at Low Loadings. ACS Catal 2013, 3, 1545– 1554.
(9) Zhang, J.; Chen, J.; Ren, J.; Sun, Y. Chemical Treatment of γ-Al2O3 and its Influence on the Properties of Co-based Catalysts for Fischer–Tropsch Synthesis. Appl. Catal. A Gen. 2003,
243, 121–133.
(10) Abi Aad, J.; Courty, P.; Decottignies, D.; Michau, M.; Diehl, F.; Carrier, X.; Marceau, E. Inhibition by Inorganic Dopants of γ-Alumina Chemical Weathering under Hydrothermal Conditions: Identification of Reactive Sites and their Influence in Fischer-Tropsch Synthesis.
ChemCatChem 2017, 9, 2106–2117.
(11) Catalysis by Transition Metal Sulphides. From Molecular Theory to Industrial
Application; Toulhoat, H.; Raybaud, P., Eds.; Editions Technip: Paris, 2013.
(12) Sarrazin, P.; Kasztelan, S.; Payen, E.; Bonnelle, J. P.; Grimblot, J. Interaction of Oxomolybdenum Species with γ-alumina and γ-alumina Modified by Silicon. 2. The molybdena/γ-alumina and molybdena/silica/γ-alumina Systems. The Journal of Physical
Chemistry 1993, 97, 5954–5961.
(13) Tougerti, A.; Llorens, I.; D'Acapito, F.; Fonda, E.; Hazemann, J. L.; Joly, Y.; Thiaudière, D.; Che, M.; Carrier, X. Surface Science Approach to the Solid–Liquid Interface:
Surface-Dependent Precipitation of Ni(OH)2 on α-Al2O3 Surfaces. Angew. Chem. Int. Ed. 2012, 51, 7697–7701.
(14) Tougerti, A.; Methivier, C.; Cristol, S.; Tielens, F.; Che, M.; Carrier, X. Structure of Clean and Hydrated α-Al2O3 (1-102) Surfaces: Implication on Surface Charge. Phys. Chem. Chem.
Phys. 2011, 13, 6531–6543.
(15) Naoto Koizumi; Takehisa Mochizuki; Muneyoshi Yamada. Preparation of Highly Active Catalysts for Ultra-Clean Fuels. Catal. Today 2009, 141, 34–42.
(16) Costa, V.; Guichard, B.; Digne, M.; Legens, C.; Lecour, P.; Marchand, K.; Raybaud, P.; Krebs, E.; Geantet, C. A rational interpretation of improved catalytic performances of additive-impregnated dried CoMo hydrotreating catalysts: a combined theoretical and experimental study. Catal. Sci. Technol. 2013, 3, 140–151.
(18) Jean-Marie, A.; Griboval-Constant, A.; Khodakov, A. Y.; Diehl, F. Influence of Sub-Stoichiometric Sorbitol Addition Modes on the Structure and Catalytic Performance of Alumina-supported Cobalt Fischer–Tropsch Catalysts. Catal. Today 2011, 171, 180–185.
(19) Borg, Ø.; Dietzel, Pascal D. C.; Spjelkavik, A. I.; Tveten, E. Z.; Walmsley, J. C.; Diplas, S.; Eri, S.; Holmen, A.; Rytter, E. Fischer–Tropsch Synthesis: Cobalt Particle Size and Support Effects on Intrinsic Activity and Product Distribution. J. Catal. 2008, 259, 161–164.
(20) Bonduelle-Skrzypcak, A.; Guichard, B. Where Coordination Chemistry Meets HDT Catalyst Synthesis Organic Additives. In Catalysis by Transition Metal Sulphides. From
Molecular Theory to Industrial Application; Toulhoat, H., Raybaud, P., Eds.; Editions Technip:
Paris, 2013; pp 181–208.
(21) Dumond F.; Marceau, E.; Che, M. A Study of Cobalt Speciation in Co/Al2O3 Catalysts Prepared from Solutions of Cobalt−Ethylenediamine Complexes. J. Phys. Chem. C 2007, 111, 4780–4789.
(22) Jiao, W. Q.; Wang, Y. M.; He, M.-Y. Morphology-controlled Synthesis of γ-Al2O3 with Large Mesopores Through Combustion of Aluminum Carboxylate Salts. Microporous
Mesoporous Mater. 2013, 181, 123–131.
(23) Stanislaus, A.; Al-Dolama, K.; Absi-Halabi, M. Preparation of a Large Pore Alumina-based HDM Catalyst by Hydrothermal Treatment and Studies on Pore Enlargement Mechanism.
Fifteenth Indian National Symposium on Catalysis and Second Conference of the Indo-Pacific Catalysis Association 2002, 181, 33–39.
(24) Absi-Halabi, M.; Stanislaus, A.; Al-Zaid, H. Effect of Acidic and Basic Vapors on Pore Size Distribution of Alumina Under Hydrothermal Conditions. Appl. Catal. A Gen. 1993, 101, 117– 128.
(25) ZHANG Jun-Ling, CHEN Jian-Gang, DONG Qing-Nian, REN Jie, SUN Yu-Han. Adsorption and Reaction Behavior of FT Co Catalysts Supported by Chemically-modified Alumina. Chem.
J. CHINESE U. 2003, 24, 301–304.
(26) Ravenelle, R. M.; Copeland, J. R.; Kim, W.-G.; Crittenden, J. C.; Sievers, C. Structural Changes of γ-Al2O3-Supported Catalysts in Hot Liquid Water. ACS Catal 2011, 1, 552–561.
(27) Koichumanova, K.; Sai Sankar Gupta, K. B.; Lefferts, L.; Mojet, B. L.; Seshan, K. An in situ ATR-IR Spectroscopy Study of Aluminas Under Aqueous Phase Reforming Conditions. Phys.
Chem. Chem. Phys. 2015, 17, 23795–23804.
